Diagnose Why a Prompt Keeps Producing Generic Output

You are a prompt engineer performing a diagnosis. Work in stages and do not rewrite anything until stage 3.

MY MATERIALS
The prompt I used: [PASTE CURRENT PROMPT]
What it produced: [PASTE OUTPUT, or describe it]
What I actually wanted: [DESIRED OUTCOME]
The tool I'm using: [CHATGPT / CLAUDE / GEMINI / OTHER]

STAGE 1 — DIAGNOSE
Categorize every problem you find into:
- Missing context the model could not have known
- Ambiguity, where two readings are both valid
- Internal conflict, where two instructions fight each other
- Unstated success criteria, where I never said what good looks like
- Wrong format specification, or none at all
- Over-constraint, where I strangled the useful output
Quote the exact fragment causing each problem. If the output was generic, name the specific missing input that caused it — do not answer "add more detail".

STAGE 2 — EXPLAIN
In 3 sentences, state the single root cause. Most weak prompts have one, not six.

STAGE 3 — REBUILD
Rewrite the prompt. Include only the components this task actually needs — role, context, audience, goal, inputs, constraints, output format, missing-information handling, quality check. Justify in one line why you included or excluded each. Do not pad it into a giant template.

STAGE 4 — TEST
Run the rebuilt prompt mentally and show a short sample of the output it would now produce. If the sample is still generic, fix the prompt again and say what you changed.

STAGE 5 — HAND OFF
Give me:
1. The final prompt, clean and copy-ready.
2. The 3 questions you would still need answered to make it materially better.
3. One reusable lesson about my prompting habits, based on what you found.